Traditional Chinese medicine knowledge graph generation method and system based on large language model

Abstract

The invention provides a traditional Chinese medicine knowledge graph generation method and system based on a large language model. The method comprises the following steps: S1, constructing a model; s2, inputting data; s3, acquiring conditions; s4, outputting data; s5, feedback is obtained; s6, case recording; s7, updating the model; existing data and real-time data are fed to the learning training model for deep learning calculation, so that an auxiliary treatment scheme is provided for a doctor according to patient information during treatment, the treatment scheme and associated data are visually displayed together, the doctor can conveniently acquire data according to requirements, and the treatment efficiency is improved. A treatment scheme is repeatedly optimized in a feedback rating mode, so that the treatment effective rate is improved, and the quality of the treatment scheme is improved; according to the system, various key information of a patient is input through the data input unit, so that the effective rate of a treatment scheme is improved, and existing cases are compared through the online data module when the key information is input into the deep learning unit.

Description

Technical Field

[0001] The present invention relates to a knowledge graph generation method and system, and in particular to a traditional Chinese medicine knowledge graph generation method and system based on a large language model, belonging to the field of traditional Chinese medicine treatment auxiliary technology. Background Art

[0002] Knowledge graphs are different from ordinary knowledge bases. Knowledge graphs integrate all disciplines, linking knowledge units of different sources, types, and structures into graphs. Based on the metadata of each discipline, they provide users with a broader and deeper knowledge system that is constantly expanding. In essence, they systematize and relate domain knowledge data and visualize knowledge.

[0003] Traditional Chinese medicine has a lot of related knowledge, and manual data retrieval is inefficient. Visualizing traditional Chinese medicine knowledge through knowledge graphs can effectively improve data acquisition efficiency. However, during the treatment process, simple knowledge graphs only call fixed data and cannot provide doctors with treatment plan recommendations based on diagnosis cases. Treatment mostly relies on the doctor's experience, resulting in unstable treatment effects.

[0004] In order to solve the above technical problems, a method and system for generating a traditional Chinese medicine knowledge graph based on a large language model is proposed. Summary of the invention

[0005] In view of this, the present invention provides a method and system for generating a traditional Chinese medicine knowledge graph based on a large language model to solve or alleviate the technical problems existing in the prior art and at least provide a beneficial option.

[0006] The technical solution of the present invention is implemented as follows: A method for generating a TCM knowledge graph based on a large language model comprises the following steps:

[0007] S1. Build a model, and build a deep learning model based on TCM drug data and diagnosis and treatment data;

[0008] S2. Input data and use diagnosis and treatment examples as training data to feed the deep learning model, so that the model has preliminary TCM knowledge response capabilities;

[0009] S3, acquisition conditions, input key information of patient treatment into the model;

[0010] S4, data output, the model integrates data according to patient information, and outputs the integrated data and related data in a visual way;

[0011] S5. Obtain feedback. After using the data integrated by the model as a treatment plan, rate and feedback the model based on the treatment effect.

[0012] S6, case records, use the treatment cases as deep learning data to continue feeding the model for optimization;

[0013] S7. Model update: perform regular maintenance and update of the model, and archive the model after each update.

[0014] Further preferably, in S1, when constructing the model, a rating feedback function is added to the model, and the priority of the scheme provided by the model is determined according to the rating feedback after treatment. Each treatment scheme includes at least three alternative schemes, and when outputting the scheme, the main scheme and the three alternative schemes are output together.

[0015] Further preferably, in S2, when data is entered for model learning, manual supervision is required to check for messy data, and when the model has response capabilities, continuous learning optimization and response capability testing of the model are performed.

[0016] Further preferably, in S3, the information input into the model needs to include data such as the patient's physical condition, the patient's medical record, age, medications taken and medications that cannot be taken.

[0017] Further preferably, in said S4, when outputting data, related data other than the treatment plan are displayed in an abbreviated form, and extended information is provided when necessary to ensure the simplicity of the visualized data.

[0018] A TCM knowledge graph generation system based on a large language model, comprising a deep learning unit, a data input unit and a data output unit;

[0019] The deep learning unit includes a learning and training module, a feedback and rating module, and a data integration module; the data input unit includes a body detection module, a voice input module, and an online data module; and the data output unit includes a visualization module, a manipulation module, and a rating module.

[0020] Further preferably, the feedback rating module rates the treatment plans according to the treatment feedback, and abolishes ineffective plans after rating, and enters effective plans into the learning and training module for use as effective data.

[0021] Further preferably, the body detection module detects the patient's body data, obtains the patient's body data, and cooperates with the voice input module to obtain other relevant information. The voice input module is equipped with an invalid word filtering function to record and convert valid information in the voice.

[0022] Further preferably, the visualization module displays the integrated data, and the manipulation module can manipulate the data displayed by the visualization module, thereby displaying related data.

[0023] The embodiment of the present invention has the following advantages due to the adoption of the above technical solution:

[0024] 1. The present invention feeds existing data and real-time data into the learning and training model for deep learning calculation, so as to provide doctors with auxiliary treatment plans according to patient information during treatment, and visualize the treatment plans and related data together, so as to facilitate doctors to obtain data according to needs, and repeatedly optimize the treatment plans through feedback rating, thereby improving the treatment efficiency and the quality of the treatment plans;

[0025] 2. The present invention inputs various key information of the patient through the data input unit, thereby improving the effectiveness of the treatment plan. The more information entered, the more accurate the treatment plan. When entering the deep learning unit, the existing cases are compared through the online data module, thereby improving the stability of the treatment plan.

[0026] The above summary is for illustrative purposes only and is not intended to be limiting in any way. In addition to the illustrative aspects, embodiments and features described above, further aspects, embodiments and features of the present invention will be readily apparent by reference to the accompanying drawings and the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0028] Figure 1 A method flow chart of a method for generating a TCM knowledge graph based on a large language model in the present invention;

[0029] Figure 2 This is a functional module diagram of a TCM knowledge graph generation system based on a large language model in the present invention.

[0030] Attached figure numbers: 10, deep learning unit; 11, learning and training module; 12, feedback and rating module; 13, data integration module; 20, data input unit; 21, body detection module; 22, voice input module; 23, online data module; 30, data output unit; 31, visualization module; 32, manipulation module; 33, rating module. DETAILED DESCRIPTION

[0031] In the following, only some exemplary embodiments are briefly described. As those skilled in the art will appreciate, the described embodiments may be modified in various ways without departing from the spirit or scope of the present invention. Therefore, the drawings and descriptions are considered to be exemplary and non-restrictive in nature.

[0032] The embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0033] like Figure 1 As shown, the embodiment of the present invention provides a method for generating a TCM knowledge graph based on a large language model, comprising the following steps:

[0034] S1. Build a model, and build a deep learning model based on TCM drug data and diagnosis and treatment data;

[0035] S2. Input data and use diagnosis and treatment examples as training data to feed the deep learning model, so that the model has preliminary TCM knowledge response capabilities;

[0036] S3, acquisition conditions, input key information of patient treatment into the model;

[0037] S4, data output, the model integrates data according to patient information, and outputs the integrated data and related data in a visual way;

[0038] S5. Obtain feedback. After using the data integrated by the model as a treatment plan, rate and feedback the model based on the treatment effect.

[0039] S6, case records, use the treatment cases as deep learning data to continue feeding the model for optimization;

[0040] S7. Model update: perform regular maintenance and update of the model, and archive the model after each update.

[0041] In one embodiment, in S1, when constructing the model, a rating feedback function is added to the model. After treatment, the priority of the plan provided by the model is determined based on the rating feedback. Each treatment plan includes at least three alternative plans. When outputting the plan, the main plan and the three alternative plans are output together.

[0042] In one embodiment, in S2, when data is entered for model learning, manual supervision is required to check for messy data, and when the model has responsiveness, continuous learning optimization and responsiveness testing are performed on the model.

[0043] In one embodiment, in S3, the information input into the model needs to include data such as the patient's physical condition, the patient's medical records, age, medications taken, and medications that cannot be taken.

[0044] In one embodiment, in S4, when outputting data, related data other than the treatment plan is displayed in an abbreviated form, and extended information is provided when necessary to ensure the simplicity of the visualized data.

[0045] like Figure 2 As shown, the embodiment of the present invention also provides a TCM knowledge graph generation system based on a large language model, comprising a deep learning unit 10, a data input unit 20 and a data output unit 30;

[0046] The deep learning unit 10 includes a learning and training module 11, a feedback and rating module 12 and a data integration module 13; the data input unit 20 includes a body detection module 21, a voice input module 22 and an online data module 23; the data output unit 30 includes a visualization module 31, a manipulation module 32 and a rating module 33.

[0047] In one embodiment, the feedback rating module 12 rates the treatment plans according to the treatment feedback, and after rating, abolishes the ineffective plans, and enters the effective plans into the learning and training module 11 for use as effective data.

[0048] In one embodiment, the body detection module 21 detects the patient's body data, obtains the patient's body data, and cooperates with the voice input module 22 to obtain other relevant information. The voice input module 22 is equipped with an invalid word filtering function to record and convert valid information in the voice.

[0049] In one embodiment, the visualization module 31 displays the integrated data, and the manipulation module 32 can manipulate the data displayed by the visualization module 31 to display related data.

[0050] When the present invention is working: after constructing the model, relevant data is fed to the model for deep learning training. After the deep learning unit 10 completes preliminary learning, the deep learning unit 10 is manually tested to ensure the accuracy of the plan provided by the deep learning unit 10. During treatment, the patient's physical data is obtained through the body detection module 21, and other information is entered through the voice input module 22. Invalid words are filtered out when entering voice information. According to the obtained information, the deep learning unit 10 retrieves and integrates the relevant data to obtain the corresponding treatment plan and related data, and the treatment plan and related data are visually output and displayed. The related data is displayed in an abbreviated manner to ensure the simplicity of the data during display. The abbreviated data is expanded through the control module 32. After the treatment is carried out using the treatment plan, feedback is returned to the deep learning unit 10 according to the treatment results, the plan is rated according to the feedback, invalid plans are deleted, and the effective plan is used as learning data to feed the model to optimize the deep learning unit 10.

[0051] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of various changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for generating a traditional Chinese medicine knowledge graph based on a large language model, characterized in that: The following steps are involved: S1. Build a model, and build a deep learning model based on TCM drug data and diagnosis and treatment data; S2. Input data and use diagnosis and treatment examples as training data to feed the deep learning model, so that the model has preliminary TCM knowledge response capabilities; S3, acquisition conditions, input key information of patient treatment into the model; S4, data output, the model integrates data according to patient information, and outputs the integrated data and related data in a visual way; S5. Obtain feedback. After using the data integrated by the model as a treatment plan, rate and feedback the model based on the treatment effect. S6, case records, use the treatment cases as deep learning data to continue feeding the model for optimization; S7. Model update: perform regular maintenance and update of the model, and archive the model after each update.

2. According to the method for generating a TCM knowledge graph based on a large language model according to claim 1, it is characterized in that: In S1, when constructing the model, a rating feedback function is added to the model. After treatment, the priority of the plan provided by the model is determined based on the rating feedback. Each treatment plan includes at least three alternative plans. When outputting the plan, the main plan and the three alternative plans are output together.

3. According to the method for generating a TCM knowledge graph based on a large language model according to claim 1, it is characterized in that: In S2, when data is entered for model learning, manual supervision is required to check for messy data, and when the model has response capabilities, the model is continuously optimized for learning and tested for its response capabilities.

4. The method for generating a TCM knowledge graph based on a large language model according to claim 1, characterized in that: In S3, the information input into the model needs to include data such as the patient's physical condition, the patient's medical records, age, medications taken and medications that cannot be taken.

5. According to claim 1, a method for generating a TCM knowledge graph based on a large language model is characterized in that: In said S4, when outputting data, the associated data other than the treatment plan is displayed in an abbreviated form, and extended information is provided when necessary to ensure the simplicity of the visualized data.

6. A TCM knowledge graph generation system based on a large language model, characterized by: It comprises a deep learning unit (10), a data input unit (20) and a data output unit (30); The deep learning unit (10) comprises a learning training module (11), a feedback rating module (12) and a data integration module (13); the data input unit (20) comprises a body detection module (21), a voice input module (22) and an online data module (23); and the data output unit (30) comprises a visualization module (31), a manipulation module (32) and a rating module (33).

7. A TCM knowledge graph generation system based on a large language model according to claim 6, characterized in that: The feedback rating module (12) rates the treatment plans according to the treatment feedback, and after rating, it abolishes the ineffective plans and enters the effective plans into the learning and training module (11) for use as effective data.

8. According to claim 6, a TCM knowledge graph generation system based on a large language model is characterized by: The body detection module (21) detects the patient's body data, obtains the patient's body data, and cooperates with the voice input module (22) to obtain other relevant information. The voice input module (22) is equipped with an invalid word filtering function to record and convert valid information in the voice.

9. According to claim 6, a TCM knowledge graph generation system based on a large language model is characterized by: The visualization module (31) displays the integrated data, and the manipulation module (32) can manipulate the data displayed by the visualization module (31), thereby displaying related data.

Images